Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
|
from
login
1.
Improving Fuzzy Robustness in LLM Safety with Automated Red Teaming
(
alignment.openai.com
)
1 point
by
Klaster_1
8 days ago
|
past
|
discuss
2.
GPT-6 Astra covertly underperforms on safety research
(
alignment.openai.com
)
342 points
by
lucrbvi
16 days ago
|
past
|
196 comments
3.
Eval-production behavior gaps in agent safety and the role of eval-awareness
(
alignment.openai.com
)
2 points
by
Klaster_1
17 days ago
|
past
4.
Beyond g-mean²: complementary metrics for monitorability
(
alignment.openai.com
)
1 point
by
gmays
28 days ago
|
past
5.
OpenAI models disable misalignment monitoring and suppress alerts
(
alignment.openai.com
)
1 point
by
rphv
31 days ago
|
past
6.
Suppressing misalignment monitoring alerts and disabling misalignment monitoring
(
alignment.openai.com
)
2 points
by
armcat
31 days ago
|
past
7.
Suppressing misalignment monitoring alerts and disabling misalignment monitoring
(
alignment.openai.com
)
5 points
by
dnw
31 days ago
|
past
8.
Beyond g-mean²: complementary metrics for monitorability
(
alignment.openai.com
)
1 point
by
Klaster_1
31 days ago
|
past
9.
OpenAI AI Model Hacked Government Website While Counting Hospital Beds
(
alignment.openai.com
)
7 points
by
DeepLogin
34 days ago
|
past
|
2 comments
10.
Using a funded cryptocurrency wallet without authorization
(
alignment.openai.com
)
2 points
by
pamcake
35 days ago
|
past
11.
Using a funded test cryptocurrency wallet without authorization
(
alignment.openai.com
)
3 points
by
ben_w
36 days ago
|
past
12.
Studying metagaming latents in language models
(
alignment.openai.com
)
1 point
by
Klaster_1
44 days ago
|
past
13.
Auto-review of agent actions without synchronous human oversight
(
alignment.openai.com
)
4 points
by
tosh
46 days ago
|
past
14.
Command injecting a reference tool to copy a source file
(
alignment.openai.com
)
1 point
by
droidjj
47 days ago
|
past
15.
OpenAI Misalignment Reports and Notices
(
alignment.openai.com
)
2 points
by
fzimmermann89
53 days ago
|
past
16.
Self-replicating prompt injections exist
(
alignment.openai.com
)
2 points
by
jonbaer
55 days ago
|
past
17.
Exposing a GitHub token in a public repository
(
alignment.openai.com
)
3 points
by
mplappert
55 days ago
|
past
|
4 comments
18.
OpenAI: Self-replicating prompt injections exist
(
alignment.openai.com
)
2 points
by
dnw
55 days ago
|
past
19.
An OpenAI agent used DNS to reach an external chatbot
(
alignment.openai.com
)
16 points
by
Metacelsus
56 days ago
|
past
|
1 comment
20.
An agent used DNS to reach an external chatbot
(
alignment.openai.com
)
198 points
by
apsec112
56 days ago
|
past
|
188 comments
21.
OpenAI Misalignment Reports
(
alignment.openai.com
)
1 point
by
macleginn
64 days ago
|
past
22.
Uploading Files to the Internet in Order to Cite Them
(
alignment.openai.com
)
19 points
by
derbOac
65 days ago
|
past
|
5 comments
23.
OpenAI models secretly generate instructions to ignore constraints
(
alignment.openai.com
)
125 points
by
theahura
65 days ago
|
past
|
37 comments
24.
Encouraging Deception in Compaction Summaries
(
alignment.openai.com
)
2 points
by
aesthesia
65 days ago
|
past
25.
Measuring reward-seeking by instilling contrastive beliefs
(
alignment.openai.com
)
11 points
by
mfiguiere
4 months ago
|
past
|
1 comment
26.
Reinforcement learning towards broadly and persistently beneficial models
(
alignment.openai.com
)
2 points
by
spicypete
4 months ago
|
past
27.
Reinforcement learning towards broadly and persistently beneficial models
(
alignment.openai.com
)
2 points
by
gmays
4 months ago
|
past
28.
Reinforcement learning towards broadly and persistently beneficial models
(
alignment.openai.com
)
2 points
by
vesteny77
5 months ago
|
past
29.
Reinforcement learning towards broadly and persistently beneficial models
(
alignment.openai.com
)
1 point
by
jawiggins
5 months ago
|
past
30.
OpenAI: Investigating the consequences of accidentally grading CoT during RL
(
alignment.openai.com
)
2 points
by
pretext
6 months ago
|
past
More
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search: